Challenge 22.1

Challenge 22.1 – Predicting NPS Classification


Customers of TracFone Wireless, Inc. draw from diverse and vibrant backgrounds. For example, they may be proudly pragmatic who appreciate good value mobile service plans. Alternatively, they may be travelers that are temporarily visiting friends or family or forming new memories on their vacation to take home with them. They may even belong to financially fragile communities that have traditionally been under served by the large wireless providers that required credit checks and long-term contracts. Ever since our inception in 1996, our focus has always been on providing Coverage & Access For All

TracFone was a pioneer in the USA market to offer mobile phones and plans without extensive credit requirements, making smartphones accessible to a wider part of the community.

The Net Promoter Score (Reichheld, Frederick F. (December 2003). “One Number You Need to Grow”. Harvard Business Review. 81 (12): 46–54, 124.) is a widely used key performance indicator to gauge customer loyalty, satisfaction, and enthusiasm toward a brand, a product, or a service. Nowadays, Net Promoter Score (NPS) is used by almost every business to evaluate customers’ sentiment. At TracFone, we measure NPS to monitor customers’ satisfaction on the overall services and products we offer. NPS, for each customer, is calculated from responses to ad-hoc questions in surveys (“Based on your experiences as a customer, on a scale from 0-10, where 0 is not at all likely and 10 is extremely likely, how likely are you to recommend our brand to a friend or colleague?”). Based on customers’ NPS scores, each customer can be classified into one of three classes: Promoter, Passive, and Detractor. Although many customers respond to questionnaires, many others do not respond. How can we determine to which class non-responding customers belong?


The 22.1 challenge is to use machine learning to predict the NPS class of non-responding customers. This is a three class predictor problem: Promoter, Passive, and Detractor.

Customer data contain various aspects of customer interactions with our services and products. Can data points, once featured engineered, be used to predict to which NPS class each customer belongs? If we can accurately predict the NPS class, we can pinpoint customers who are dissatisfied with our services. By predicting dissatisfied customers, we can then work with them to closely understand their pain points, and develop solutions to improve their experience with TracFone.


1st place –

  • TBA

2nd place –

  • TBA

3rd place –

  • TBA


Who can join in?

TracHack 22.1 participants will be students enrolled in the Master of Science in Business Analytics (MSBA) in A/Prof Daniel McGibney’s class.

When & where?

TracHack 22.1 is an online competition. The data for the competition is released April 11, 2022 and ends May 1, 2022 at 11:59 pm US Eastern Time.

How are the winners selected?

Teams are welcome to make as many submissions along the way, and the accuracy of solution will be evaluated based on F1 score. The winners of the challenge will be announced based on the ranking from the final submission. That means your team needs to ensure your final submission is the best solution that you come up with during the three-week period.

The top three teams with the highest F1 score will win cash prizes, fame and glory. Even though the final ranking is based on the final submission, it is a good idea to validate the approach and solution along the way using regular submissions as checkpoints.

What technologies can I use?

TracHack 22.1 solutions must be built using PySpark. TracFone will provide a dedicated environment with PySpark and Jupyter, dedicated for each team. Teams will access data, build their solutions and submit their predictions via that dedicated environment. So it is important for you to get familiar with it and set it up once your team signs up and gets the relevant details. Keep in mind that the winning teams will have to submit the code that produced their winning solution once the winners are announced for verification prior to the awards.

Where is the data?

The data will be released on April 11 accessible from within that environment. Teams may not download part or all of the data locally. Teams downloading the data outside that environment will be disqualified from the competition. All data and code developed by the teams must remain in the team’s environment.

How do I submit my solution?

The details are available on the submissions page and consult the FAQ page.

What happens when my team wins?

You win fame, glory and bragging rights. There are also cash prizes for the teams to go with trophies:

1st place team: $3,000

2nd place team: $2,000

3rd place team: $1,000

The winning teams will be required to make a 5-minute presentation at the award ceremony on May 4 that describes their solution and how they tackled this problem.

How do I get started?

  • Understand the Rules and consult the FAQ.
  • Join a team (if not already) – Check with A/Prof Daniel McGibney.
  • Sign an NDA on data usage – If you are part of a signed up team you would have gotten a DocuSign email.
  • Ensure you are set up to access your team’s dedicated PySpark development environment. – Check your emails for setup and login instructions once you have signed the NDA.
  • Set up to make your Submissions.

Key dates

March 4, 2022TracHack 22.2 Problem Announced
April 11, 2022TracHack 22.2 Data Released
May 1, 2022 – 11:59pm US ETDeadline for Final Submission
May 4, 2022 – 3:30am US ETAwards & Presentations


If you have any questions, feel free to reach out to us at Once your team is registered you’ll be invited to a Microsoft Teams Channel.